serious test bench
 

http://www.hydrocool.com/Testapparatus.html

there is an engr, one Herr Montag, who has posted on OCAU
- perhaps he will visit

good gear and all, but am at a loss to understand how equilibrium is established after only 15 min. ?

also note the effect of a 144x144mm heat die (see the OCAU thread)

Ben, your business model is looking more and more remote
price that setup, then look to the personnel

Wow. And I was thinking how hot the new 3497 was with the new actuator outputs and T t-c input card.... These guys really seem to know how to do it right. So how soon do you get the precision temp controlled cave, Bill? Gotta keep up with the joneses, right? Do you think they might leave the chiller and cave running all the time, and then temp control the die simulator with the mounted sensor so it's all close to equilibrium to start with? It still would seem to take longer than 15 minutes to let everything settle, but you gotta admire good engineering.

14.6*14.6... Only close to an Athlon 64, and they don't have support for A64's, making their results useless...

Would be interesting to ask him whether this was obtained on their 14.6 mm sq (213.16 mm^2) die.
I note the gif is now removed from their "list of performance graphs", have copy here if link doesn't work.
However this comparison still remains on their list.

This doesn't seem consistent with my own testing as far as RBX vs.Cascade goes

Herr Montag mentioned 14.9x14.9mm over at OCAU for a 222mm^2 die.

Perhaps another reason why not to trust any manufacturer's own comparisons of competing products?

Just focusing on a product that's not my own, being the RBX in this instance.

At ~0.105 C/W for the RBX at 1GPM for a 222mm^2 heat die, that certainly does not seem to fit well with results seen for JoeC's 130mm^2 die, or even really match up that well if we extrapolate to BillA's test-bed.

In fact, the results for both the RBX and the Cascade seem particularly "whack", and given the independent tests of the block that are available, do not agree at all with Hydrocool's findings.

Joec's results would certainly indicate that given a pump like the MCP600 that the RBX would be in front once we take the relative pressure-drop of the respective blocks into account and the impact that will have on the flow rate that the pump could push through either block.

Am having trouble correlating the apparant quality of the testbed system to the results that are coming out the other end...

Noticed that. Additionally my memory says their site originally referred to a 146 mm^2 die (but memory can play tricks and prob was only a typo)
Will wait for more info.

It's possible the RBX is being tested with the #5 nozzle plate plus the sensor is located in the center (assuming it's the same testbench described). I'm not sure were the cpu sensor is located but I'd think an off center location would be more benefical to the Cascade in a comparison between the two. I can't see any reason for a manufacture to falsely make one block better than another when they're both not his.

Or is JoeC's 140 mm^2 ?
To quote Bill, "JoeC told me yesterday the die area was 140mm²
sorry Les, should have posted" - link

Would also like to know the offset (from TIM interface) of the die temperature sensor.
Bill's was ~ 2mm, but has this been reduced by lapping?
Awaiting more info on die sizes before spending too much time on sums.

I have a very strong suspicion that the Cascade in question was the same user-modified Cascade that WCP tested, which also under-performed, presumably due to the user modifications done.

methinks our 'usage' of "C/W" may be inadaquate as the area is undefined

should it not be °C/W/cm² ?
(or whatever area over which the determination has been made ?)

Yes. Although this is implied by the tester when the test system is documented.

°C/W/cm² is not linear with varying area though. Comparitively greater thermal spread with small areas than larger areas. Results are still specific to the test-bed die-size.

but look how often the documentation is lacking and/or incomplete
my preference is to include the area as an essential part of the 'units' description

size differences recognized, yet another reason to be including the area

No argument there from me. The test die-size, and its applicability to the problem at hand, is of critical importance when C/W's get thrown around.

let's hope we can avoid an l/d discussion
- I have NO data (given a sq heat die, and as are the Intel ttvs - have an IHS)

Cathar
you visit more forums than I (only 2), promote the inclusion of the heat source area in the units of thermal resistance
- if you are so inclined

I'll start adding it to the Swiftech descriptions

I pretty much always attempt to source and quote the die-size when mentioning C/W's.

Of more benefit to the community would be respected testbed setups that document the die sizes along with a few paragraphs on the importance of doing so. Phaestus and JoeC would be the two best places to start. JoeC's test-bed documentation could certainly do with a bit more fleshing out, if only to explain just why the phrase "test results are not comparible with others" appears on his results page.

Would go a longer way towards wider educational benefit than either you or I posting in forums, and potentially being viewed somewhat cynically by various forum members for being "tainted" with the manufacturer brush.

Edit: Off to bed again for my second 4 hrs sleep. I hate jet lag.

15 minutes doesn't seem realistic to me either, especially with a 150W load (takes longer). A temp log would dissipate any doubts.

Actually, I've been looking at this flowmeter:
http://cgi.ebay.com/ws/eBayISAPI.dll...MEDW%3AIT&rd=1

Massive, but with 0.15% accuracy, and .10% repeatability (and other factors). Still in negotiations. I'm beginning to realize that pressure measurement is cheaper.

°C/W/cm² wouldn't work for certain reasons. While a block may do good at a low area, it may fall behind with higher areas (a Cascade wouldn't do very well with a 50*50mm area, but a MCW6000 would remain very strong, for example) With °C/W and a defined size (yes, this does need to be added) the results will be constant. Increase the wattage by 1000 times and the value remains the same, but this won't happen with increase the surface area by 1000 times.

not intended as a "solution" to anything,
just as a relevant piece of information which should be added - or otherwise defined